Power amplifiers are widely used for example in radio base stations and user equipments in wireless communication systems. Power amplifiers typically amplify input signals of high frequencies into an output signal ready for radio transmission. High efficiency and linearity are generally desirable for power amplifiers to reduce power consumption and minimize errors and/or distortions in the output signal.
Known error or distortion correction techniques for power amplifiers include feedback, pre-distortion and feedforward techniques. Of these, only so called adaptive pre-distortion and feedforward techniques are viable for systems having wide bandwidth and with tough linearity requirements.
Adaptive pre-distortion, typically used in digital implementations, is a linearization technique that works by providing an inversely nonlinear signal to the input of a nonlinear amplifier so that the output signal becomes linear. To shape the nonlinear input signal to the amplifier, the adaptive pre-distortion technique uses sampling of the amplifier's output signal together with nonlinear modeling and adaptive signal processing. A major benefit of this technique is that the efficiency of the amplifier is almost unaffected.
However, the adaptive digital pre-distortion technique cannot counteract noise and handles several types of distortion poorly, or not at all. The pre-distortion signal generally has much higher bandwidths than the final output signal, especially for compression, low or negative gain slope regions and sharp kinks in the transfer function. Digital pre-distortion systems need a correct set of model parameters, which sometimes is hard to determine. A specific set of model parameters might not work in practice if the produced amplifiers behave differently than the model. The signal processing complexity, and consequentially the size and power consumption, can be high for complex error processes. These problems are exacerbated by requirements for high bandwidths and low distortion.
Feedforward, e.g. described in Seidel, H., “A microwave feed-forward experiment,” Bell System Tech. J., pp. 2879-29 16, Nov. 1971, is a linearization technique that works by injecting a corrective signal after a main amplifier A1 to restore linearity, as shown in FIG. 1. With this method error extraction is done by a first signal sampling coupler C3 shown in FIG. 1. An amplified output signal from the main amplifier A1 is sampled by a signal sampling coupler C2, the first signal sampling coupler C3 then compares the sampled signal with a reference input signal IN and outputs an error signal. The reference input signal IN is delayed by a transmission line or a delay filter L1 in order to be in sync with the amplified output signal from the main amplifier A1. The error signal is then amplified by an error amplifier A2 to a corrective or compensation signal and injected by C4 to an output OUT. A delay line L2 after the main amplifier A1 ensures that the corrective, or compensation, signal is injected in sync with the output signal from the main amplifier A1. The box marked with X may be an inversion, or inverter, in the cases where the main amplifier A1 is a non-inverting amplifier. Feedforward systems are often described in terms of “loops”, i.e. an error extraction loop followed by an error injection loop.
Error injection performed by element C4 in FIG. 1 is handled by either a transformer or a directional coupler. The directional coupler has the advantage that it has high backwards isolation, i.e. the injected signal mainly goes forward to the output, whereas the transformer sends half of the injected signal power back towards the main amplifier A1.
Due to limited precision in gains, phases and delays, two or more feedforward stages are usually required to reduce the error to specified levels even if the loops are adaptively adjusted. A feedforward stage is also commonly used to complement a pre-distortion system to handle the “difficult” types of errors.
Feedforward method can handle any type of errors, e.g. noise, gain, frequency response variations and all types of distortions including nonlinear memory effects with arbitrary time constants, negative gain regions and even hysteresis. It can do this at high frequencies, over wide bandwidths and without knowledge of the specific error processes involved. It therefore has advantages over pre-distortion techniques both when it comes to the types of errors that can be handled and bandwidth. Since feedforward method also corrects for noise in the main amplifier path, high-selectivity high-power filters after the main amplifier may be eliminated in a well-designed feedforward system. The noise and distortion requirements on the pre-distortion part of a combined pre-distortion-feedforward system may also be relaxed.
However, a drawback of the feedforward method is low efficiency. This is to a large extent due to losses in the error injection coupler and low efficiency of the error amplifier. Usually, large losses in the error injection coupler and low efficiency in the error amplifier will result if the maximum error signal, either in voltage or current, needs to be handled is large. These losses behave differently for transformer couplers and directional couplers. Other losses and inefficiency come from the delay line after the main amplifier, the signal sampling couplers, and limited precision in loop balancing with respect to gains, phases and delays. This means that the error amplifiers must have headroom to accommodate residual signal instead of only the error signal.
Transformer coupling has no specific coupler loss, but influences the efficiency of the error amplifier. It has larger loss for low amplitude signal in the main path. That will lower efficiency for the error amplifier, since the error amplifier is not isolated from it. Large loss also comes from sending half of the injected signal power in the wrong direction. The part of the injected signal that goes backwards reflects at and interacts with the main amplifier which gives rise to new distortion products and ripples in the output signal, which is known as interaction problems. Due to these problems the transformer coupling method has been substantially abandoned in favor for the directional coupler method.
A directional coupler has close to zero coupling loss for injected signals that are in phase with and proportional, by a coupling factor, to the output signal from the main amplifier, but has high loss for injected signals that are far away from these conditions. Even if the error in the output signal is zero, the directional coupler dumps part of the output signal power from the main amplifier into a resistance. It however does not have a specific penalty for signals with low amplitude in the main path, since the error amplifier is isolated from it by the directional coupler. Furthermore, the average efficiency of the error amplifier is low if the error in the output signal is small on average compared to the maximum error.
→3a